In just under a year — since OpenAI released its now-household-name ChatGPT onto the world — generative artificial intelligence (genAI) has incomparably shifted the attention of enterprises large and small.

Simply put, the world-changing technology is set to revolutionize the way organizations run their businesses (not to mention possibly change life as humanity knows it).

But enterprises are increasingly realizing the challenges in existing, widely used “horizontal AI” — or broad-reaching AI that can be layered across a multitude of industries and applied to numerous use cases.

This is where “vertical AI” is coming into play. Many call this emerging technology the next step forward in AI and vertical software-as-a-service (vSaaS), or software that’s tailor-made for specific industries.

“While horizontal AI solutions still hold value in providing more generalized capabilities, vertical AI solutions are expected to provide more value in the future as organizations begin to rely more heavily on generative AI,” Shahar Chen, CEO and cofounder of enterprise AI software vendor Aquant, told SDxCentral.

Horizontal AI: Versatile but limited at the same time

Horizontal AI is what many describe as the first wave of AI and the technology as most of us know it today. It is a generalized, applicable, foundational technology that can be applied across industries and address a diversity of business needs.

Examples include inescapable industry game-changers such as ChatGPT, Google’s Bard, Salesforce’s Einstein, and Anthropic’s Claude.

“The adaptability of [horizontal AI] solutions makes them highly accessible for businesses seeking AI that can be swiftly customized to meet their specific requirements,” said Chen.

However, he noted, “horizontal AI solutions may not offer the same depth of domain-specific knowledge and expertise as their vertical counterparts, limiting their ability to deliver accurate and personalized insights.”

The technology provides generic outputs that “lack nuance and specialty” and there’s also a competition component, Chen said. With horizontal AI swiftly becoming commonplace, enterprises might find it challenging to differentiate themselves using off-the-shelf tools.

Industry-specific and tailored

This is where vertical AI is said to excel. The technology is industry-specific and tailored — examples include Darktrace for cybersecurity, AlphaSense for finance, Jasper for marketing or Harvey for law firms.

“Vertical AI solutions are designed to provide targeted and specialized functionalities,” said Chen. These tools leverage industry-specific knowledge and expertise to offer more targeted, personalized and reliable results. Specialized algorithms are specifically designed for an industry or use case and are often offered as plug-and-play because they are typically prebuilt.

They are also designed with adherence to industry-specific regulations and compliance standards.

“Unlike horizontal AI, which aims to provide general-purpose solutions, vertical AI is tailored to specific industry verticals or niches,” said Jeff Mains, CEO of Champion Leadership Group, which helps B2B SaaS founders and CEOs scale their companies. “The key differentiator lies in its deep domain expertise.”

As a result, vertical AI can improve user experience and lower computing costs because it focuses on a specific training set that requires fewer parameters, thus requiring less computing power for inferencing, Chen pointed out.

Furthermore, these tools can offer more data security: Lower resource requirements mean enterprises can often run models on their own cloud infrastructure and share their proprietary data without getting third parties involved.

Demand for narrow and specific

Ultimately, the technology is nothing less than a “significant evolution” and a “game-changer” in the world of AI, according to Mains.

“Firstly, it can significantly enhance operational efficiency by automating industry-specific tasks and processes,” he said. “Secondly, it enables better decision-making through data-driven insights tailored to the unique challenges of the industry.”

Vertical AI can also offer personalized customer experiences, which are “increasingly critical in today's market,” he pointed out. Industries such as healthcare, finance, manufacturing and retail stand to benefit the most due to their complex and specialized needs.

“The rise of vertical AI brings precision, efficiency and competitiveness to enterprises by addressing specific industry challenges head-on,” said Mains. He further emphasized that it's crucial for enterprises that want to gain the most benefit from the technology to collaborate with AI providers who understand the intricacies of their market.

Paris Heymann, partner and investor at Index Ventures, agreed, calling vSaaS part of a broader trend of end users increasingly demanding tailored technology products.

“Consumers want solutions-oriented software made specifically to solve their exact business problems,” he wrote. “In an environment where we are inundated with software, narrow and specific is well-positioned versus broad and generalized.”

Challenges of vertical AI

Still, as with any technology, vertical AI has its challenges, too.

It can take a little longer to get to market than horizontal AI because product builders have to put together specific market datasets that refine outputs of underlying foundation models. They then typically require reinforcement learning from human feedback (RLHF) to train apps for customer expectations.

Vertical AI can also have versatility and generalization limitations, said Hannah Overhiser of Prophia, which provides AI and machine learning (ML) tools specifically for commercial real estate (CRE). The technology may struggle to generalize knowledge and skills to new or unfamiliar situations. A lack of adaptability limits the technology’s ability to apply knowledge beyond its intended scope.

“Modifying or repurposing a vertical AI system for a different application may require significant retraining or redevelopment, making it less agile compared to more generalized approaches,” Overhiser said.

Additionally, it is knowledge- and data-dependent, requiring access to large, representative datasets for a target domain. Acquiring and curating that data can be challenging in areas where data availability is limited, thus reducing vertical AI’s benefits. Furthermore, systems trained on domain-specific data may be susceptible to bias in training data, and their narrow focus may result in an issue called overfitting. This happens when an AI system “becomes overly tuned to the specific training data, compromising its ability to handle variations or edge cases that differ from the training distribution,” according to Overhiser.

Enterprises identifying the best AI for their needs

Benefits and challenges aside, all this is not to say that enterprises may use one technology exclusively over the other.

Before investing in any genAI — horizontal or vertical — Chen advises organizations to determine exactly what they want to achieve and how they plan to measure the success of the technology. They should also make sure that they have data readily available to power the model.

Conduct a thorough analysis of business processes, customer needs and market trends to identify use cases, he says. AI should improve efficiency, enhance decision-making and create new opportunities.

Establishing internal expertise is also important in building — and realizing — an AI strategy. Enterprises with the resources can directly hire AI specialists, or there’s always upskilling of existing employees and partnering with external experts and consultants.

When first getting started, Chen advised, perform small-scale pilot projects that test feasibility and potential impacts and challenges associated with implementing AI in the real world.

Ultimately, he noted, organizations will likely use different types of AI together to “benefit from the broad capabilities of horizontal AI and the specialized functionalities of vertical AI.”